Optimal Decision Trees in Gatekeeping Procedures for Confirmatory Clinical Trials

• Activity Number: 307
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): JonDavid Sparks*+ and Alex Dmitrienko and Haiyuan Zhu
• Companies: Eli Lilly and Company and Eli Lilly and Company and Forest Research Institute
• Keywords: multiple testing ; gatekeeping procedures ; grouping objectives

Abstract

Confirmatory clinical trials often consist of many objectives where the need to adjust for multiplicity is required. The objectives are often grouped into families where serial or parallel gatekeeping procedures may be utilized. The decision on how to group and order families (e.g., primary, secondary) may be based on clinical relevance, assumed statistical power, or for various other reasons. However, concerns can arise as the number of hypotheses in a family or the number of families increases. This may lead to a substantial decrease in power for some promising objectives or objectives may simply not be tested due to the performance in previous families. In this research, we explore approaches that group objectives into different families and test them using general gatekeeping procedures. A variety of multiple testing procedures in combination with different family groupings will be considered. The power performances of the strategies are evaluated under various nominal power scenarios. An optimal decision tree is presented which can aid researchers in developing multiple testing methods under assumed nominal power levels.